存储墙问题是高性能计算研究的核心技术问题，固态硬盘作为一种新型存储器，具有访问速度快，容量大，功耗低等优点，被认为是解决存储墙问题的一个有效途径而得到广泛应用。然而，固态硬盘也有读写开销不对称，写放大等缺点，其性能的发挥依赖于有效的固件和操作系统调度算法的性能。目前国际国内的相关研究，主要面向固件层的调度算法，在操作系统层面，尤其面向高性能计算应用的研究还不够，算法的性能还不够高，还有很多待研究的问题。因此，基于固态硬盘加速的虚存cache管理技术研究，具有重要的理论和实际应用价值。基于此，我们拟在如下方面开展研究：(1) 面向固态硬盘加速的热数据跟踪，能够为调度算法提供更充分的调度信息；(2) 基于固态硬盘的虚存管理优化，对数据预读、写回进行优化，能够捕获更多的数据复用，提高应用性能；(3) 面向分布式文件系统基于固态硬盘加速的优化，能够提高数据写、元数据服务等的性能。

Memory wall problem is the major concern to be solved in high performance computing research. Solid-state drive, as a new type of storage, has advantages in access speed, capacity and power. It is regarded as an effective solution to memory wall problem, and is widely used. However, SSD also has some shortcomings such as asymmetric read/write cost, write amplify. It relies on effective scheduling implemented in firmware as well as operating system to achieve good performance. The state of the art researches mostly focus on the firmware side, with insufficient attention on operation system, especially towards high performance computing. The scheduling performance does not meet the expectation, and there are many problems awaited to be improved. Therefore, research on SSD based virtual memory cache management is very important. We plan to conduct research on the following sides, (1) Research on hot data tracking for SSD acceleration, to provide more information for virtual memory scheduling optimization; (2) Research on SSD based virtual memory management optimization, focus on data read ahead, write back, to capture more data reuse, improve application's performance; (3) Research on SSD based distributed file system optimization, to improve data write as well as metadata service's performance.